Relational Online Application Processing or ROLAP seemed like an inefficient way to do Business Intelligence when MicroStrategy got started in 1989!
Here is an interesting demo from MicroStrategy showing how ROLAP can do things on the fly:
However increases in processor, memory, and hard disk speeds and steep fall in unit prices of those are finally proving that their approach may be very timely after all!
ROLAP has always taken the rap for a long time that calculating certain components of Business Intelligence, as and when needed, may be a very inefficient way to do Business Intelligence.
MOLAP (Multi-dimensional OLAP) on the other hand, as found in Cognos, Business Objects, etc created multi-dimensional cubes and kept them all updated as and when transactions are processed.
For example, let's say you have Point of Sale data and you need different reports ordered by different orders of analysis like:
- Product Sales by SKU, Category and Product Line
- Product Sales by SKU, Zone, Region and Division
- Product Sales by SKU, Sales Person, Sales Manager, Regional Manager
Typical ROLAP approaches as followed by MicroStrategy could compute these from the original POS database at run-time. Of course, in some cases, you may need to create temporary tables when you have huge numbers of transactions even in MicroStrategy, but nothing compared to MOLAP ones.
MOLAP approaches would create different dimensions like SKU, Category, Product Line, Zone, Sales Person, Sales Manager, etc and keep all of their totals updated in real-time, all the time. Of Course, some of them may have hierarchies within the dimensions also. SKU-Category-ProductLine, StoreLocation-City-Zone-Region-Division could be another. They require their own totals also. If the dimensions increase, computing time may also increase exponentially. Updating the cubes, and querying the cubes, could all get more and more complex with the addition of every additional dimension.
MOLAP made more sense when computing was much more expensive.
Now if you can have 1000 times the processing power at 100th the cost as compared to 1989, ROLAP makes more sense!
Now there is an even more compelling need NOT to create complicated, pre-designed, somewhat inflexible, data warehouses with many, many dimensions, since a lot of the users these days want flexibility in their reporting!
Users want to specify reports as and when they need them and additional ones to be created quickly.
This is where ROLAP truly comes into fore - flexible, on-demand reporting!
Another area wher ROLAP increasingly is very good is handling many data sources from many heterogenous environments where you don't know upfront what data you are getting and what the structure of the data would be.
Every corporate environment is dealing with this issue with increasing integration between information systems of various companies - My Logistics being integrated with UPS or Fedex. Or healthcare claims information being passed on electronically between various related companies.
How do you do BI when your data structures are not known ahead of time and changes from period to period?
ROLAP may be a better answer. Now you can see why a lot of companies may have created huge data warehouses that by the time, they are ready, are already obsolete!
MicroStrategy may have had the right answer for a long time!
Now is a gift. That's why it's called the Present - Anonymous
Leave a comment